One approach to reduce greenhouse gas emissions is lightweight design. Lightweight designs minimize material usage and weight while maintaining the required functionality (for example in [1]: the compliance of a lightweight turbine blade). In the case of fiber-reinforced plastic structures, even greater savings can be anticipated when spatially varying the fiber volume content and fiber orientation in line with the expected loads. However, such load- dedicated graded fiber architectures greatly expand the design space. To cope with this increased complexity, predictive simulations and eventually even automatic optimizations are desired. This presentation will discuss how the components of an automatic design cycle (design parameterization, performance evaluation, automatic optimization) can be realized to achieve load-dedicated graded designs of lightweight fiber-reinforced plastic structures. A particular focus will be on how the complexity of the application limits the viability of conventional approaches and where machine learning might help. For example, an ideal geometric design parameterization would be a single parameterized CAD model that can instantiate all geometries of interest. However, it is unclear how to derive such a parameterization in case of geometric designs that exhibit substantially different features. This challenge might be addressed by Deep Local Shapes [2], a machine-learning-based approach to geometry representation. REFERENCES [1] P. Antolin, M. Bartoň, G.-P. Bonneau, A. Buffa, A. Calleja-Ochoa, G. Elber, S. Elgeti, G.G. Escudero, A. Gonzalez, H.G. Barrio, S. Hahmann, T. Hirschler, Q.Y. Hong, K. Key, M.-S. Kim, M. Kofler, N. Lopez de Lacalle, S. de la Maza, K. Rajain, and J. Zwar. On design, analysis, and hybrid manufacturing of microstructured blade-like geometries, Computer-Aided Design, Vol. 190, 103967, 2026. [2] R. Chabra, J.E Lenssen, E. Ilg, T. Schmidt, J. Straub, S. Lovegrove, and R. Newcombe. Deep local shapes: learning local SDF priors for detailed 3D reconstruction, arXiv [cs.CV], 2003.10983, 2020.